Add a per-inode lock around btrfs_drop_extents

btrfs_drop_extents is always called with a range lock held on the inode.
But, it may operate on extents outside that range as it drops and splits
them.

This patch adds a per-inode mutex that is held while calling
btrfs_drop_extents and while inserting new extents into the tree.  It
prevents races from two procs working against adjacent ranges in the tree.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

4 files changed, 21 insertions, 0 deletions

diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h
index 81c0444f37b6..3bf40591742a 100644
--- a/fs/btrfs/btrfs_inode.h
+++ b/fs/btrfs/btrfs_inode.h
@@ -32,6 +32,7 @@ struct btrfs_inode {
         struct extent_io_tree io_tree;
         struct extent_io_tree io_failure_tree;
         struct mutex csum_mutex;
+        struct mutex extent_mutex;
         struct inode vfs_inode;
         struct btrfs_ordered_inode_tree ordered_tree;
 
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 343d1101c31c..4036c62b6671 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -1671,6 +1671,8 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
         ret = __free_extent(trans, root, bytenr, num_bytes, root_objectid,
                             ref_generation, owner_objectid, owner_offset,
                             pin, pin == 0);
+
+        finish_current_insert(trans, root->fs_info->extent_root);
         pending_ret = del_pending_extents(trans, root->fs_info->extent_root);
         return ret ? ret : pending_ret;
 }
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 3e4e5c227c0c..40ad1b2958cb 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -242,6 +242,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans,
         u64 end_of_last_block;
         u64 end_pos = pos + write_bytes;
         u64 inline_size;
+        int did_inline = 0;
         loff_t isize = i_size_read(inode);
 
         start_pos = pos & ~((u64)root->sectorsize - 1);
@@ -275,6 +276,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans,
                 if (hole_size > 0) {
                         btrfs_wait_ordered_range(inode, last_pos_in_file,
                                                  last_pos_in_file + hole_size);
+                        mutex_lock(&BTRFS_I(inode)->extent_mutex);
                         err = btrfs_drop_extents(trans, root, inode,
                                                  last_pos_in_file,
                                                  last_pos_in_file + hole_size,
@@ -289,6 +291,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans,
                                                        0, 0, hole_size, 0);
                         btrfs_drop_extent_cache(inode, last_pos_in_file,
                                         last_pos_in_file + hole_size -1);
+                        mutex_unlock(&BTRFS_I(inode)->extent_mutex);
                         btrfs_check_file(root, inode);
                 }
                 if (err)
@@ -321,6 +324,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans,
                 /* step one, delete the existing extents in this range */
                 aligned_end = (pos + write_bytes + root->sectorsize - 1) &
                         ~((u64)root->sectorsize - 1);
+                mutex_lock(&BTRFS_I(inode)->extent_mutex);
                 err = btrfs_drop_extents(trans, root, inode, start_pos,
                                          aligned_end, aligned_end, &hint_byte);
                 if (err)
@@ -332,9 +336,13 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans,
                                            inline_size, pages, 0, num_pages);
                 btrfs_drop_extent_cache(inode, start_pos, aligned_end - 1);
                 BUG_ON(err);
+                mutex_unlock(&BTRFS_I(inode)->extent_mutex);
+                did_inline = 1;
         }
         if (end_pos > isize) {
                 i_size_write(inode, end_pos);
+                if (did_inline)
+                        BTRFS_I(inode)->disk_i_size = end_pos;
                 btrfs_update_inode(trans, root, inode);
         }
 failed:
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 293355c92a4f..0d1e2ba47a9c 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -516,6 +516,8 @@ int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end,
                                           trans->transid, inode->i_ino,
                                           ordered_extent->file_offset, &ins);
         BUG_ON(ret);
+
+        mutex_lock(&BTRFS_I(inode)->extent_mutex);
         ret = btrfs_drop_extents(trans, root, inode,
                                  ordered_extent->file_offset,
                                  ordered_extent->file_offset +
@@ -531,6 +533,8 @@ int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end,
         btrfs_drop_extent_cache(inode, ordered_extent->file_offset,
                                 ordered_extent->file_offset +
                                 ordered_extent->len - 1);
+        mutex_unlock(&BTRFS_I(inode)->extent_mutex);
+
         inode->i_blocks += ordered_extent->len >> 9;
         unlock_extent(io_tree, ordered_extent->file_offset,
                     ordered_extent->file_offset + ordered_extent->len - 1,
@@ -1399,6 +1403,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
 
                 trans = btrfs_start_transaction(root, 1);
                 btrfs_set_trans_block_group(trans, inode);
+                mutex_lock(&BTRFS_I(inode)->extent_mutex);
                 err = btrfs_drop_extents(trans, root, inode,
                                          hole_start, block_end, hole_start,
                                          &alloc_hint);
@@ -1412,6 +1417,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
                                                 (u64)-1);
                         btrfs_check_file(root, inode);
                 }
+                mutex_unlock(&BTRFS_I(inode)->extent_mutex);
                 btrfs_end_transaction(trans, root);
                 unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
                 if (err)
@@ -1578,6 +1584,7 @@ static int btrfs_init_locked_inode(struct inode *inode, void *p)
                              inode->i_mapping, GFP_NOFS);
         btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
         mutex_init(&BTRFS_I(inode)->csum_mutex);
+        mutex_init(&BTRFS_I(inode)->extent_mutex);
         return 0;
 }
 
@@ -1880,6 +1887,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
                              inode->i_mapping, GFP_NOFS);
         btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
         mutex_init(&BTRFS_I(inode)->csum_mutex);
+        mutex_init(&BTRFS_I(inode)->extent_mutex);
         BTRFS_I(inode)->delalloc_bytes = 0;
         BTRFS_I(inode)->disk_i_size = 0;
         BTRFS_I(inode)->root = root;
@@ -2105,6 +2113,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry,
                 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
                                      inode->i_mapping, GFP_NOFS);
                 mutex_init(&BTRFS_I(inode)->csum_mutex);
+                mutex_init(&BTRFS_I(inode)->extent_mutex);
                 BTRFS_I(inode)->delalloc_bytes = 0;
                 BTRFS_I(inode)->disk_i_size = 0;
                 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
@@ -3079,6 +3088,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
                 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
                                      inode->i_mapping, GFP_NOFS);
                 mutex_init(&BTRFS_I(inode)->csum_mutex);
+                mutex_init(&BTRFS_I(inode)->extent_mutex);
                 BTRFS_I(inode)->delalloc_bytes = 0;
                 BTRFS_I(inode)->disk_i_size = 0;
                 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;